From computing power foundations to industrial applications, the value focus of AI investment is shifting.
TECHUB NEWS Hong Kong Report｜Reporter Alma Li｜August 25, 2026
Main points: Wu Shichun summarizes AI investments into three main lines: cycle as the anchor, hard technology as the core, and policy as the momentum. He predicts that the period from 2026 to 2030 is the golden window for strategic layout in the new Kondratiev wave cycle; the essence of the Sino-US technological competition is the struggle for computing power hegemony, "the tighter the blockade, the greater the space for domestic alternatives"; the evaluation criteria for AI investments are shifting from parameters and rankings to revenue, renewal fees, and replicable business outcomes.
On August 24, the "Financial Capital Meets Sci-Tech Unicorns" forum, hosted by the Global Value Investment Association and the Hong Kong Private Equity Association, was held in Hong Kong. Wu Shichun, founding partner of Plum Blossom Capital and vice president of the Global Value Investment Association, delivered a speech titled "Opportunities and Thoughts on AI Investment," discussing the Kondratiev wave cycle, Sino-US technological competition, the "14th Five-Year Plan," and the value transfer in the AI industry chain, while providing insights into investment opportunities for the next three years.
Starting from "Bubble": Rankings Change, Frameworks Must be Stable
In the opening of his speech, Wu Shichun used "bubble" as a starting point. He humorously noted that those who haven't boarded the train often see it as a bubble, while those who have typically feel that the value is solid. The topic is light, but it points to the most fundamental contradiction in early investments: market enthusiasm shifts rapidly, while corporate growth and industrial maturity take time.
He mentioned that the rankings in the large model field have changed almost every two or three months over the past two years. If investment judgments always follow the rankings, it is difficult to establish a stable methodology.
Wu Shichun summarized his framework into three points: cycle as the anchor, hard technology as the core, and policy as the momentum. The cycle is used to judge the position of the technological revolution, hard technology determines whether a company can form hard-to-replicate capabilities, and policy influences the direction of capital, industrial resources, and market systems allocation.
Wealth in Life Depends on Kondratiev Waves
"Wealth in life depends on Kondratiev waves." This is a phrase Wu Shichun repeatedly emphasizes. The Kondratiev wave, or Kondratiev cycle, spans about 50 to 60 years, its cycle is systematically explained by the "cycle king" Zhou Jintao, with technological revolutions being the core engine driving its movement. In a person's life, there are only three real opportunities from Kondratiev waves that change one's fate; the accumulation of wealth is essentially the capturing of cycle dividends, not merely a simple addition of labor.
So, where are we now? His judgment is: the fifth Kondratiev wave's recession period is nearing its end, and the sixth Kondratiev wave—driven by AI's intelligent revolution—is just budding; the years 2026 to 2030 are the golden period for strategic layout of this wave. He likens the present to the internet in 1995: the infrastructure construction phase has just begun, and a batch of trillion-dollar market value tech companies will emerge in the future. Compared to the internet and mobile internet, AI does not just bring efficiency tools, but a reconstruction of productivity and means of production, and this dividend period may last for decades.
At the beginning of the speech, he also spoke of an old investment of his: over a decade ago, he invested several million as an angel investor in Luo Shi Robotics, being one of its earliest angel investors; this company recently went public in Hong Kong, yielding a paper return of about 1 billion yuan. He used this transaction to illustrate the meaning of long-termism: investments are earned through an understanding of the times.
Sino-US Competition: The Essence is Computing Power Hegemony
When discussing Sino-US technological competition, Wu Shichun's expression is straightforward: the essence of this contest is the struggle for computing power hegemony.
He outlined the path of sanctions: 2022 saw a cutoff of high-end GPUs, 2023 expanded the entity list, 2024 imposed manufacturing equipment blockades, and 2025 extended to advanced packaging. However, he reached a contrary conclusion—“the tighter the blockade, the greater the space for domestic alternatives.” If high-end chips and critical equipment could flow freely, many domestic industries would not necessarily have received such dense capital input in such a short period.
Thus, he delineated three golden tracks: computing power layer—domestic GPUs, computing power scheduling, intelligent computing centers; manufacturing layer—advanced packaging, Chiplet; application layer—industrial software, industry large models, AI native applications. The closer to the key bottleneck, the higher the potential technological value, but the corresponding research and development difficulty and business risks also increase.
Supporting this judgment, he summarized China's fourfold advantages: a vast manufacturing base with rich industrial scenarios, the collaborative ability of national capital, local industrial capital, and social capital, a persistent engineer dividend, and a complete industrial chain. "The Sino-US technological competition is not a short-term friction, but a marathon lasting for 20 years," he said, emphasizing that hard technology is the foundation of national competitiveness and a gift from the times for investors to traverse cycles.
Policy Provides Direction, Enterprises Still Need to Validate
Policy is the third main line in Wu Shichun's framework. The already released "Outline of the 15th Five-Year Plan for Economic and Social Development of the People's Republic of China" has dedicated a section to digital intelligent development, proposing to coordinate the advancement of computing power facilities, model algorithms, and the supply of high-quality data.
The capital market system is also increasing its adaptability to unprofitable tech companies. In June 2025, the China Securities Regulatory Commission released guidelines on setting up a growth layer in the Sci-Tech Innovation Board to enhance institutional inclusiveness and adaptability, restarting the application of the fifth set of standards for unprofitable enterprises on this board, and expanding the scope to include artificial intelligence, commercial aerospace, low-altitude economy, and other fields. On the funding side, the National Integrated Circuit Industry Investment Fund Phase III has a registered capital of 344 billion yuan, providing longer-term capital support for hard tech companies with long research and development cycles and large investment scales.
However, a clear direction does not mean that investment becomes simple. Wu Shichun likens the investment environment of the coming years to an "open-book exam": the questions are already laid out on the table, policies can highlight the key points, but enterprises still need to resolve issues of technology, product, customers, and cash flow, and investment institutions still need to independently verify, bearing the costs of erroneous judgments.
In his words, from funds, industry, demand to exits, China has constructed "the world's most certain hard technology policy closed loop."
Five-Layer Cake: The Bottom Layer Determines the Upper Limit, the Top Layer Determines Value
When discussing specific investment strategies, Wu Shichun borrowed Huang Renxun's "five-layer cake" framework: energy layer, chip layer, infrastructure layer, model layer, application layer, structured from the bottom up, each layer relies on the previous one.
"The bottom layer determines the upper limit, the top layer determines value." Energy and chips are the "hard foundation" of AI, determining the ceiling of computing power; models and applications are the "soft engine," determining commercial depth. According to his calculations, the value proportion of the chip layer is the highest, at around 35%; the infrastructure layer is about 20%; and the model layer, energy layer, and application layer each account for about 15%.
The significance of this structure is to avoid equating AI with large models. Currently, most industrial value is concentrated in the chip and model layers; as technology spreads, value will continue to flow towards computing power infrastructure and the application end. For early-stage investment institutions, greater incremental opportunities may come from the application layer, specifically products that can integrate into core business processes and enhance the efficiency of the real economy.
The Next Three Years: Three Judgments
Supply of Computing Power Will Remain Tight
Wu Shichun predicts that the tight supply of computing power may continue until 2028, with computing power scheduling and advanced packaging as core bottlenecks. Measuring computing power investments should not only consider the number of devices but also factors like energy costs, cluster utilization rates, interconnection efficiency, and software scheduling capabilities.
AI Applications Will Enter More Central Business Processes
He expects that AI applications will further shift from consumer-facing to enterprise-facing, with industrial manufacturing likely becoming the most value-concentrated scenario. Enterprises have clear demands for cost reduction, efficiency enhancement, quality control, accelerated R&D, and production safety, and products that can embed into real processes and yield measurable results are more likely to generate stable income.
Domestic Alternatives Enter Deep Water
Once the question of "whether it exists" is resolved, the market will further question "how well it works": performance, reliability, cost, and ecosystem will become new selection criteria. Companies relying solely on concepts and policy labels will be eliminated, and those that truly master technology, can achieve productization, and win customers will remain.
At the end of the speech, Wu Shichun tied the topic back to a statement on the last slide of his presentation: "A person's wealth is essentially the monetization of their understanding of the times."
Editor’s Note: This article was compiled by TECHUB NEWS reporter Alma Li based on on-site recordings and presentation materials from the forum on August 24, 2026, with omissions, and has been approved for publication by the forum organizers. The main text adopts a third-person reporting style, retaining only content that can be confirmed from on-site recordings, presentation materials, or publicly authoritative sources; unverified proper nouns, cases, and data have not been adopted. Links to publicly available government documents in this article have been verified.
Disclaimer: This article is for informational exchange only and does not constitute any investment advice.
Related Public Information:
免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。
